GDPR-compliant AI: is ChatGPT, Claude, Copilot or Gemini GDPR compliant? (2026)

Which AI tools can a European company use under the GDPR? ChatGPT, Claude, Microsoft Copilot and Gemini checked on DPA, training, EU data residency, where inference runs and the traps in each, plus a rollout checklist.

Walma Engineering·Updated 9 November 2026·12 min read
Is ChatGPT GDPR Compliant? ChatGPT vs Claude vs Copilot vs Gemini (2026)

Short answer: none of the big AI tools is GDPR compliant or non-compliant by itself; the plan and the settings decide. ChatGPT Business and Enterprise, Claude Team, Enterprise and the API, Microsoft 365 Copilot and Gemini for Google Workspace all come with a data processing addendum and keep your data out of model training. They differ on where data is stored and, above all, where the model runs. The free and personal versions of all four are consumer products and do not belong anywhere near personal data at work.

GDPR-compliant AI is not a feature you buy, it is a set of conditions a tool has to meet before you can use it lawfully with personal data. This guide checks ChatGPT, Claude, Copilot and Gemini against those conditions, shows the trap in each one, and ends with a checklist for rolling AI out in a company without your data protection officer saying no.

Facts are from the providers' own documentation, checked in November 2026. This is a practical guide, not legal advice: your DPO has the last word.

What makes an AI tool GDPR compliant?

Under the GDPR your company is the controller and the AI provider is your processor. That relationship, not the model, is what you are assessing. Seven questions cover it:

  1. Is there a data processing addendum (DPA)? Article 28 requires a contract with every processor. Consumer plans have none, which ends the discussion for personal data.
  2. Is your data used for training? If prompts can end up in the next model, you have lost control of what happens to them.
  3. Where is data stored? Chats, files and logs at rest.
  4. Where does inference run? The moment the model reads your prompt and writes the answer. This is the question most "EU hosted" claims skip: data can be stored in Frankfurt and still be processed in Virginia.
  5. Is there a lawful transfer mechanism for anything that leaves the EU: the EU-US Data Privacy Framework or standard contractual clauses.
  6. How long is data kept, and can you delete it?
  7. Can you control and audit use? Single sign-on, who has access to which model, logs you can show a regulator.

On top of that come your own duties: a legal basis for each use, transparency towards the people whose data you process, and a data protection impact assessment (DPIA) under Article 35 when the processing is likely to be high risk, which covers HR data, health data and anything that profiles people.

Is ChatGPT GDPR compliant?

It depends entirely on the plan.

  • Free, Plus and Pro are consumer products. There is no DPA, and conversations can be used to improve OpenAI's models unless the user turns that off in the data controls. Fine for private use, not for customer or employee data.
  • ChatGPT Business (formerly Team) has a DPA, and OpenAI does not train on Business content by default. Data residency in Europe is rolling out to Business workspaces, but it covers storage at rest only. See our ChatGPT Business guide for the details.
  • ChatGPT Enterprise and Edu add the stronger residency options: new workspaces can store content at rest in Europe, and since January 2026 eligible customers can also keep inference in Europe (the EEA and Switzerland). Both are agreed with OpenAI's sales team, not switched on in a menu.
  • The API and Azure OpenAI are the route for your own applications. On Azure, GPT-6.1 Sol and GPT-6 Luna run in the EU Data Zone, which includes Sweden Central; GPT-6 Astra is only offered as a global deployment, so it is not processed exclusively in the EU.

The trap: OpenAI keeps safety and abuse-monitoring data in the US even for European workspaces, including a time-limited copy of prompts and responses. It does not rule ChatGPT out, but it belongs in your DPIA.

The regulatory history is real, too: Italy's data protection authority fined OpenAI €15 million in December 2024 over how ChatGPT handled personal data, a decision OpenAI appealed. Regulators are watching this category closely. More on running the GPT models in Europe: ChatGPT in the EU.

Is Claude GDPR compliant?

  • Claude Free, Pro and Max are consumer plans. Users choose in the privacy settings whether their chats may be used to improve Claude; with that setting on, Anthropic can keep the data for up to five years. No DPA.
  • Claude Team, Enterprise and the API run on Anthropic's commercial terms with a DPA, and Anthropic does not train on that data by default. Zero data retention is available for qualifying API use.
  • Data residency is the gap. Anthropic's own plans have no EU region: as of November 2026 the only in-product residency option is US-only inference, for usage-based Enterprise. Anthropic points European customers to its cloud partners instead. On AWS Bedrock and Google Vertex AI, Claude is available in European regions with processing kept inside the EU/EEA; Microsoft Foundry is listed as "coming soon" for Europe.

The trap: "We use Claude" can mean three very different data flows: a consumer account, Anthropic's US-hosted commercial plans, or Claude in an EU cloud region. Ask which one before anyone in legal signs off. Prices for each plan are in our Claude plans comparison; running Claude in Europe is covered on Claude in the EU.

Is Microsoft Copilot GDPR compliant?

Microsoft 365 Copilot (the version tied to your company's Entra ID login) is covered by Microsoft's product terms and DPA and by the EU Data Boundary: for EU and EFTA customers, Copilot data at rest stays in the EU, apart from limited pseudonymized data Microsoft keeps for security and operations. For many companies that already run Microsoft 365, it is the easiest tool to get approved.

The trap is a setting called flex routing. When it is allowed, Copilot's language-model inference may run in the United States, Canada or Australia during periods of peak demand. Microsoft's own documentation says it is on by default for EU tenants created after 25 March 2026, and asks older tenants to check their setting. To turn it off, an AI Administrator goes to Copilot → Settings → View all → Flex routing during peak load periods and selects Do not allow flex routing.

Two more things to check: Anthropic's models inside Copilot fall outside Microsoft's EU Data Boundary commitments and are disabled by default for EU, EFTA and UK tenants, so leave them off unless your DPO has looked at them. And the consumer Copilot that comes with a personal Microsoft account is a different product with different terms.

Is Gemini GDPR compliant?

  • Gemini in a Google Workspace business or enterprise edition is covered by Google's Workspace data processing terms, and Google does not use that content to train its models. An administrator has to accept the data processing amendment in the Admin console; it does not apply automatically.
  • Data regions let admins keep Gemini data in the EU: Workspace data regions have covered Gemini features since 2025, and since June 2026 they also cover the Gemini app on Enterprise and Education editions, with EU storage and processing configurable per organizational unit.
  • The trap: data regions do not cover everything. Logs and cached content are explicitly outside them, and some features rely on services that process data globally. Read Google's list of what is covered before you promise anyone "EU only".
  • Gemini on a personal Google account is a consumer product: conversations can be reviewed by people to improve the service. Keep company data out of it.

The four tools side by side

ChatGPTClaudeMicrosoft 365 CopilotGemini (Workspace)
Business plans with a DPABusiness, Enterprise, Edu, APITeam, Enterprise, APIYes, with Entra IDBusiness and Enterprise editions
No training on your data by defaultYes on business plansYes on business plansYesYes
Data at rest in the EUEnterprise, Edu; Business rolling outNot on Anthropic's own plans; yes via Bedrock or Vertex EUYes (EU Data Boundary)Yes, with data regions
Inference in the EUEnterprise and Edu, eligible customersVia Bedrock or Vertex EU regionsYes, unless flex routing is allowedConfigurable per data region
Trap to checkSafety data kept in the USWhich of three data flows you are onFlex routing, Anthropic modelsLogs and cache outside data regions
Consumer version OK for personal data?NoNoNoNo

What about data transfers to the US?

All four providers are American, so transfers come up even when data sits in Europe. The legal basis today is the EU-US Data Privacy Framework: the EU General Court upheld it on 3 September 2025 (case T-553/23), and an appeal is pending at the Court of Justice (case C-703/25 P). Its two predecessors, Safe Harbor and Privacy Shield, were both struck down, which is why many DPOs keep standard contractual clauses as a fallback and prefer setups where data is both stored and processed in the EU. Residency does not make a US provider's obligations under US law disappear, but it does shrink what can be transferred at all.

GDPR and the AI Act are two different laws

The GDPR governs personal data; the EU AI Act governs AI systems. Meeting one does not cover the other. Most office use of ChatGPT, Claude, Copilot or Gemini is not high-risk under the AI Act, but the duty to support AI literacy among staff already applies, and AI used in recruitment or employee management falls under the high-risk rules, which apply from December 2027. Our guide to AI governance covers how the two fit into one policy.

How to roll out GDPR-compliant AI in a company

  1. Find out what is already in use. In most companies people are using personal ChatGPT and Claude accounts today. That shadow AI is your biggest GDPR exposure, and banning it rarely works; replacing it does.
  2. Pick business plans only. A DPA and no training by default are the entry ticket. Block consumer AI domains on company devices once the replacement is live.
  3. Decide where inference may run. Write it down per tool: EU only, or EU with documented exceptions. Then configure it: data residency in ChatGPT Enterprise, flex routing off in Copilot, EU data regions in Gemini, an EU region for Claude.
  4. Run a DPIA for the uses that touch sensitive or large amounts of personal data, and update your record of processing activities.
  5. Write a short AI policy that says which tools are approved, what data may go into them and who to ask. Train people on it; that is also your AI Act literacy duty.
  6. Control access and keep logs. Single sign-on, model access per team, and usage logs you can produce if a regulator or a customer asks.
  7. Review twice a year. Residency and routing options changed several times in 2026 alone: OpenAI added inference residency in January, Microsoft introduced flex routing in the spring and Google extended data regions to the Gemini app in June.

Step 3 is where companies get stuck, because every provider solves residency differently and no single plan gives you all the models in Europe. That is the problem Walma AI Hub is built for: Claude, GPT and the other leading models through one gateway hosted in the EU, with Claude via AWS Bedrock and GPT via Azure OpenAI in European regions, access controlled per team and every request logged. Models that have no EU option yet can be blocked or allowed per team, so the choice your DPO made in step 3 is enforced rather than hoped for.

Frequently asked questions

Is ChatGPT GDPR compliant?+

ChatGPT Business and Enterprise can be used in a GDPR-compliant way: both come with a data processing addendum (DPA) and OpenAI does not train on their content by default. Enterprise and Edu can keep data at rest in Europe and, for eligible customers, run inference in Europe too. Business residency covers storage only. The free version, Plus and Pro are consumer products without a DPA and are not suitable for personal data at work.

Is Claude GDPR compliant?+

Claude Team, Enterprise and the API come with Anthropic's commercial terms and a DPA, and Anthropic does not train on that data by default. Anthropic's own plans have no EU data region, so companies that need processing in the EU run Claude through AWS Bedrock or Google Vertex AI in European regions, or through a gateway hosted in the EU. Claude Free, Pro and Max are consumer plans without a DPA.

Is Microsoft Copilot GDPR compliant?+

Microsoft 365 Copilot is covered by Microsoft's DPA and the EU Data Boundary, so data at rest stays in the EU. Check one setting: flex routing lets inference run in the United States, Canada or Australia at peak load, and it is on by default for EU tenants created after 25 March 2026. Admins can turn it off under Copilot, Settings, Flex routing during peak load periods.

Is Gemini GDPR compliant?+

Gemini in a Google Workspace business edition is covered by the Workspace data processing terms and is not used to train Google's models. Admins can set data regions so that Gemini data is stored and processed in the EU, but logs and cached content are not covered by data regions. Gemini on a personal Google account is a consumer product and not suitable for company data.

Which AI is GDPR compliant?+

No AI tool is GDPR compliant on its own: compliance depends on the contract, the settings and how you use it. What you need is a business tier with a DPA, no training on your data, a clear answer on where data is stored and processed, admin controls and logging, and a DPIA where the processing is high risk. ChatGPT Enterprise, Claude via an EU cloud region, Microsoft 365 Copilot with flex routing off and Gemini for Workspace with EU data regions can all meet that bar.

Can I put personal data into ChatGPT?+

Not into the free version, Plus or Pro: there is no DPA and conversations can be used to improve the models. In ChatGPT Business or Enterprise you can, provided you have a legal basis for the processing, have told the people concerned, and have assessed the risk. Special category data such as health data needs extra care and usually a DPIA.

Is the EU-US Data Privacy Framework still valid?+

Yes. The EU General Court upheld the adequacy decision on 3 September 2025 (case T-553/23, Latombe). An appeal is pending at the Court of Justice (C-703/25 P), so many companies keep standard contractual clauses as a fallback and prefer providers that can process data inside the EU.

Walma AI Hub

The same tools, in your EU region, under your control

A 20-minute walkthrough with an engineer. We map it to your tools, your MCP servers and your budget model.

About AI Hub